https://www.techpowerup.com/gpudb/2871/radeon-rx-vega
GPU Name: Greenland
GPU Variant: Greenland XT
Architecture: GCN 1.3
Clock: 1200 MHz
Bandwidth: 512 GB/s
225 W
Lol no
This is complete horseshit
>>60113135
>28th April
Already ordered mine! SHould be here yesterday
>>60113135
>GCN 1.3
AHAHAHAHAHAHAHA
>completely new arch
where are the dumb poojeets now
>>60113135
Fake news
So it released yesterday and nobody cared? AMD BTFO
So AMD lied to me again! I was planning on spending $2000 on a new AMD building but not anymore I can't taake their lies and deciet anymore
>>60113135
>Raja's face when
>>60113170
>>60113135
Great work TPU, this is underage twitch shooter youtube-gaymur-tier reporting.
>>60113135
Fake news
TPU is such fucking garbage.
>>60113329
>TDP: 250W
>>60113186
Uhh no one said it was a completely new arch.
If you have to make up things other said for your shitposts, you already lost.
>>60113351
Seems normal for a highend card?
Also nobody really knows the TDP, it's speculation at the moment.
>below the 1080
lol
>>60113390
Another failure as always
ANOTHER
MASSIVE
DISAPPOINTMENT
I FUCKING KNEW IT VEGA WAS SHIT HAHAHAHAHAHAHAHAHAHAHAHAHAH AMDTARDS BTFO BTFO BTFO BTFO BTFO BTFO BTFO BTFO BTFO BTFO DIE POO SCUM DIE POO SCUM DIE POO SCUM DIE POO SCUM DIE POO SCUM DIE POO SCUM DIE POO SCUM
TPU is baiting now, wew. I knew they were mediocre, but falling this low.
The only thing they got right there is Bus Interface and 14nm
Maybe the power connectors too?
Lmao
YES YES NVIDIA BTFOS AMD AGAIN! GO BANKRUPT ALREADY YOU SHITTY POO COMPANY!!!
POO POO POO POO POO POO
SHITTY POOGA GOT BTFO TAKE THAT FUCKING POO SHITTERS!
>>60113329
>TPU is such fucking garbage.
this.
TPU is a known nvidia shill site.
And they are still salty that they didn't get free Fury samples.
>>60113135
i didn't know Salazar Studio got hired by TPU
>>60113459
Even that fuckshit admitted the R5's killed the i5's in his last video, it took out his soul to admit it though, just like with Stephen Burke.
>>60113355
>no one said it was a completely new arch
oh boy the lying poojet backtrack cycle
like clockwork
>>60113474
>Even that fuckshit admitted the R5's killed the i5's in his last video
good, anything to drop this dudes ego down a few notches is good
>>60113329
>>60113456
that's what you get when a nazi in in charge of a site and a pooinloo is in charge of it's """content"""
>>60113390
It's "estimated" performance. If it's actually true and Vega is actually below a stock 1080 from almost a year ago AMD is fucking done and the high-end GPU market will get to "enjoy" hardcore NVIDIA assrape.
>>60113390
That performance graph would put the card at the same performance as the Fury X.
fake as fuck.
>>60113390
If this is real I don't regret my 1070.
>>60113604
It's real!
>>60113604
>If this is real I don't regret my 1070.
Your 1070 comes with telemetry in the drivers.
>>60113600
>Fury X is 5% behind a 1080
top lmao
>>60113604
It's absolutely 100% real, it can't be anything else, they just got most of the specs wrong but that's just a minor mistake
Has shitposting gone too far?
>>60113647
It is in some titles.
Are Nvidiots so technologically incompetent that some actually believe these specs?
>>60113646
Yes, as AMD also does. I'm not sure if my 2016 driver has already telemetry.
>>60113646
At least you can delete it from the installer and the driver with its control panel still work fine.
>>60113714
AMD has ads in the control panel that you can turn off.
Nvidia puts stealth spyware on the computer and you have to hack files out of the installer to disable it, and then the drivers still end up bricking your card.
>>60113135
"Radeon RX Vega performance estimated based on architecture, shader count and clocks."
Into_the_trash.jpg
>>60113135
>1200MHz clocks
When the Enterprise Vega has 1540MHz?
TPU is fucking clueless.
>>60113775
They probably didn't get any free samples this time either, so they are posting more FUD in retaliation, like when the Fury launched.
>>60113479
Uhh no sweetie
>>60113679
Yeah and in some other titles 1080 is 50% faster
what the fuck is your point, pajeet?
>>60113355
>Uhh no one said it was a completely new arch.
IT
NEVER
FUCKING
ENDS
Jesus christ I am laughing so hard right now. EVERY SINGLE TIME amd announces something, you guys go overhyping through the roof and then when info starts coming out you all backpedal on everything you said and start with this horseshit that "nn-no one said it w-would be good".
This is utterly fucking amazing.
>>60114811
>>60114811
other anon and TPU or whatever are retarded.
Vega isn't as big a change as TeraScale VLIW -> GCN SIMT, but it's a complete overhaul in a way that none of the prior GCN 1.x updates have been.
Vega will still be fundamentally a collection of up to 4 Shader Engines, each with Compute Unit blocks made of 4, 16*fp32 SIMT execution units that have coherent L1 caches, but that's about it.
Cache coherent ROPs, tiling/binning rasterizers, and 2x rate fp16 will boost performance, but they're pretty compartmentalized upgrades. The move to virtualized local memory a.k.a. "high bandwidth cache controller" a.k.a. your HBM pool is now a huge L3 cache will probably be its biggest differentiating feature, provided that they actually get any software support.
>>60114986
If serious performance to be gained, support will definitely come.
Nvidia can't even beat the AMD in gameworks anymore.
>>60114986
Honestly, all those changes to the compute blocks are a bit unnecessary. They just need to fix the fucking rasterizers so they aren't a huge bottleneck anymore. They managed to do something about it with Polaris (which is as fast as the 290x despite half the ROPs), if Vega does that even better, then they'll already have a kickass card.
>>60117783
That would be the 'fast' way. They decided to take their time and literally redo frontend from scratch.
>>60117852
Yeah, we can only hope that that doesn't end up with so many bugs that the card ends up another Fury.
Hell, I wouldn't be surprised if the card is already complete, but they are holding back the release because it takes so long to build up an inventory due to the shit yields of 14nm + large chip + 2x HBM + interposer.
>>60117908
Fury was basically double Tonga. This is something VERY different.
>shit yields of 14nm + large chip + 2x HBM + interposer.
14nm LPP yields are excellent.
>>60117357
>create a boogeyman software than apparently brings AMD cards to it's knees
>HAHA LOOK THE AMD CARD PERFORMS BETTER IN GIMPWORKS
Have you ever stopped to think that maybe it's because the boogeyman you created is wrong?
>>60117908
The silicon was finished before december, they've spent the last months building inventory , improving drivers (80% of all the driver teams are woking on Vega) and raising clocks
>>60113135
2 years, new node just add one TF bullshit.
>>60117943
Tapeout was December, but that's not final silicon revision.
>>60113135
Isn't Vega supposed to be 12,5+ tflops instead of 9,8? This would place Vega around the 1080ti.
>>60118032
Yes. TPU are a bunch of fucking retards.
TPU is officially as bad as WCCFTech
>>60118043
>>60118062
TPU isn't bad, they are just NVidia shills.
>>60118079
Who cares, that article is fucking retarded.
>>60118079
Even WCCFTech doesn't ignore previous official specs before publishing bullshit rumors
TPU just did
I like wfftech for the shitposting in the comments. Its like the yahoo news of the comp world
>GCN 1.3
didnt know vega was polaris
>>60118377
Now you know, buy a 1080
>>60114713
>ramesh calling others pajeet
lol
>>60118689
nice shilling pajeet
>>60118377
It's NCU
>>60118778
Internally it's called GCN5.0
https://en.wikipedia.org/wiki/Graphics_Core_Next#GCN_5th_Generation_.28Vega.29
NCU is either the marketing name or the new shader name
>>60118825
The architecture itself is called GFX9 internally
https://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-LLVM-GFX9
>>60114811
>I have to act so dumbfounded by your logic that people don't notice I have no argument!
>>60114811
This.
>>60113135
i invested in amd for my economics and now im going to fail because i up $9000 virtual monies on them
i cant fail i need vega to be as good as at least maxwell
>poozen
>pooga
>pooMD
What a filthy "company". That's what happens when you have a streetshitter CEO.
>>60113646
I simply use a custom configured foss firewall on the router level.
>>60113775
so they bench a card at 1200mhz and bash it saying it cant beat a 1080 yet confidently leave out stocck clock speed is going to be closer to 1600 making it on par with the 1080ti...
so witch is it? is this real or fake! if its real all the morons crying "HAHHA poor amdonres" just proved it will blow away the 1080
>>60113600
The Fury X is on the same fucking graph retard.
>>60113714
>>60113753
The botnet comes with the Nvidia driver itself now, not the installer anymore.
The worst AMD did was put a fucking link on the desktop, and they backed down on it. All that shit can be removed/turned off.
>>60113390
Barely. For half the price, you Inteljewnigger
Why AMD not cut some FP64 units to reduce chip consumption and focus on performance in other areas?
>>60117934
>Have you ever stopped to think that maybe it's because the boogeyman you created is wrong?
Of course it is wrong to create such software. But nvidia was caught with their pants down during the Crysis 2 era (the rendering an invisible ocean deal) and have since consistently been found to try and gimp the competitions performance by running mostly unnecessary (i.e. there is no visual improvement) shit that only their cards are able to process efficiently.
This has been known for a fact and is undeniable.
What is also undeniable that after some time, AMD's cards catch up and when their drivers allow the OS to actually use the more broad performance of AMD cards, they fuck nvidia's gameworks shit up so much that even nvidia's own cards cannot catch up anymore, because hardware-wise, they are worse cards but have (or had) a driver advantage
>>60113135
Is that April 29th 2018 or 2019?
>>60113135
Whoever made that pic fucked up by not using "Simultaneous Multi Tracking".
>>60124326
there was one like that
>>60113390
>Estimated boost clock 1200 Mhz
>NCU is optimized for higher clock speeds
What seems wrong here?
>>60122825
IIRC Vega does not have dedicated FP64 units
>>60113135
Just thought, isn't multi-track drifting weirdly apt for AMD due to their obsession with async?
>>60124399
GCN doesn't have dedicated fp64 units, period.
>>60113186
What are you up to?
Are you saying Nvidia churns out a new architecture every release just because they have a new name such as Maxwell and Pascal?
The latest major architectural change for Nvidia was Maxwell (900 Series). Before that it was Thermi (400). Even then, they aren't completely new architectures either.
If you think so, you're fucking retarded and just because AMD decided to name updates to GCN in incremental version steps doesn't mean that there aren't many underlying changes to the arch itself.
>>60124518
The market has spoken, highly efficient brute force GPUs > wimpy overheating async GPUs
AMD should get on board or get left even further behind than they already are
>>60124670
>highly efficient brute force GPUs
Brute force is actually what VLIW5 and GCN are, compared to Nvidia.
Nvidia has always had more of a brainiac architecture compared to AMD.
Also, GCN per se isn't bad at all. It had a ton of shortcomings that hindered its performance and efficiency, but some of these are now finally being addressed with Vega.
>>60118970
>streetshitter CEO
The Chinese is now street shitters?
>>60113135
>LIterally Fiji
Oh man.
>>60124518
Maxwell actualy was a new arhitecture.
Pascal wasn't, but still presented a massive leap forward in terms of performance and efficiency because of the die shrink and much higher clockspeeds enabled by it
>>60113135
>Already ordered my GTX 1080
Please be trash.
>>60124823
It will be faster than a 1080, but also significantly more expensive.
>>60113135
AMD is for low tier office shit. Novidya and Intel deliver a refined end user experience.
>>60124815
It was a major redesign, not a completely new architecture as far as I know.
Arguing about this at that point is more like arguing semantics though.
>>60124851
Define what you mean by 'a completely new architecture'
Gpu manufacturers always build upon their last architecture, you expect them to start from scratch with each new generation of cards?
>>60124834
>it will be 5-10% slower than 1080 but $100-$150 cheaper
ftfy
>>60124880
Yeah...no
We already know the die size which is 80% larger than that of the 1080, and it has HBM2.
Combine those two and you have a card that is way more expensive to manufacture.
>>60124823
Should have waited for Volta. My 1080 by MSI can do pliable 4k gaming 30-60 fps, current and older games. But it'll start struggling by the time its successor releases.
>>60124823
>Buying new GPU
>in April
You are what we call in the trade, a moron.
>>60124904
In that case I stand corrected, it will be 5-10% slower than the 1080Ti etc...
>>60124916
You can get 1080's for 300 dollars less than at launch.
I would call people who buy gpus at launch morons
>>60124904
>we know it's 80% larger
I would like citation please
>more expensive to manufacture
yea but selling price =/= manufacture price
GTX1080 is cheaper to manufacture than GTX980 yet it costs more.
>>60124866
>Gpu manufacturers always build upon their last architecture
Obviously, which is why I said arguing about this is mostly pointless.
I'd say GCN was a new, from scratch architecture, as were Pentium 4 and Zen, but those aren't GPUs.
>you expect them to start from scratch with each new generation of cards
I don't but the retard I was initially replying to implied that there can't be any changes because it's still based on and called GCN.
>>60124909
wew, are we now playing the waiting game on Nvidia's side too?
>>60124925
New cards are on horizon buying gtx1080 now is gambling. You should buy new cards in November.
>>60124930
>I would like citation please
http://wccftech.com/amd-vega-gpu-pictures-hbm2-official/
>yea but selling price =/= manufacture price
No shit, but it's still the biggest part of the selling price
>GTX1080 is cheaper to manufacture than GTX980 yet it costs more.
I would like a citation for that
>>60124944
The 1080 was still being sold for almost 850 in november, now for 550, so I don't think so.
New cards are always on the horizon and in terms of price/performance buying shit at launch is always the worst decision.
>>60124953
As there is citation to be had, I was going by your logic of smaller die.
>>60124961
No, new cards aren't always on the horizon. Next cards after Vega (which will release in two months at the very worst, rumored may release), are in 2018. That's everything but the horizon.
>>60124996
Next cards after Vega are Volta.
Q3-Q4 2017.
So yes, new cards are always on the horizon
>>60124953
>It has large ceramic, it must be GPU
They might of just enclosed HMB traces, because they were problem with Fiji.
>>60125009
People calculated the die size, excluding the HBM
Here you go
http://www.overclock.net/t/1620103/various-exclusive-amd-vega-presentation/80
It's up to you if you want to believe it or not, it's certainly in line with what AMD has been saying so far in terms of shader units.
>>60113135
Ryzen is for smooth gayming and good productivity therefore the best, Intlel is for kiddies who enjoy 20 min fps in gayming.
>>60125021
Ceramic enclosure doesn't mean that it has full die behind it.
HBM2 on that Vega chip is 2x size of the JEDEC standard.
>>60121030
How did you come to the conclusion that they benched it?
Do you even read?
>>60125046
None of what you are saying is making any sense
>>60125044
>smooth gaming
Yeah..no
>>60125080
on Fury X, there were exposed traces to HBM, that you could of moved by cleaning the chip.
>>60125088
Not my fault you can't accept reality.
>>60125095
wtfamireading.jpg
>>60125097
>my memes are facts
Ok kid
>>60125088
never seen 1% or 0.1% benchmark? :^)
>>60125098
>Retard wants to talk about Fiji
>Retard doesn't know shit about Fiji
Amazing.
>>60125106
You're talking absolute fucking horseshit, that's not my fault
>>60125103
Nice strawman, good to know you admitted that you're wrong.
>>60125007
volta is 2018
>>60125104
He's another braindead pootel shill, what did you expect
>>60125007
>Q3-Q4 2017
that's just a rumor from 1 source only
>>60113135
postin specs